lvogel123/gpqa-diamond-grok-4

该数据集分为三个配置：结果、样本和统计。每个配置都包括特定的特征，例如log_path、eval_id、run_id、created、task、task_id和model等。结果配置关注于性能指标，如准确度；样本配置包括每个样本的详细信息，如纪元、目标和消息；统计配置提供了关于模型使用和令牌统计的信息。数据集被划分为训练集，并为每个配置提供了文件路径。

The dataset is divided into three configurations: results, samples, and stats. Each configuration includes specific features such as log_path, eval_id, run_id, created, task, task_id, and model among others. The results configuration focuses on performance metrics like accuracy, while the samples configuration includes details about each sample like epoch, target, and messages. The stats configuration provides information about model usage and token statistics. The dataset is split into a training set, and file paths for each configuration are provided.